Ningjia Yu Undergraduate Dissertation 2017/18

Topic Similarity

Supervised by M.Stevenson

Abstract

Information is a new currency. With the rapid development of the internet, people are able to access to information in few second though few keystrokes and clicks. With the great convenience, there are new problems come behind. How to sort through the enormous amount of information and found the most relevant one in short period of time. To achieve this target, many types of research about the topic model and topic similarity was carried out in the past.

This project mainly focuses on different implement methods based on word probability distributions, distributional semantic methods, knowledge-based approaches etc. Topics are generated by topic model Latent Dirichlet allocation (LDA) and evaluate the results against each other and human judgement.